Merck & Co.

Merck (MSD outside the United States and Canada) discovers, makes and sells prescription medicines and vaccines, and through its Animal Health division, medicines for pets and farm animals. In 2025 it sold $65.0bn: $58.1bn of human medicines and vaccines and $6.4bn of animal health. It employs about 75,000 people and is based in Rahway, New Jersey.

One product dominates. Keytruda is a cancer immunotherapy: an antibody that takes the brakes off the patient's own immune system so it can attack tumours. Since its approval in 2014 it has been approved for dozens of uses, from lung cancer, its largest, to melanoma, kidney, bladder, breast and others, and it became the best-selling medicine in the world. In 2025 it brought in $31.7bn, 49% of Merck's sales. Merck does not publish profit by product, but a medicine of that size, made at that scale and sold at that price, probably earns well over half of the company's profit.

The second franchise is vaccines: Gardasil, which prevents the cancers caused by HPV ($5.2bn in 2025, down from $8.6bn in 2024 after sales to China stopped), childhood vaccines and Capvaxive for adult pneumonia. Then a long list of other medicines: Winrevair for a rare lung disease, Welireg for kidney cancer, shares in Lynparza and Lenvima, and older drugs like Januvia and Bridion that have just lost their patents.

Drug patents last about twenty years from filing, and the useful selling life is usually ten to fifteen. When a patent ends, cheaper copies arrive and sales of the original fall by most of their value within a few years. Every large drug company is therefore running to stand still: it must discover or buy new medicines faster than the old ones expire. Merck's problem is that the thing about to expire is half the company.

Merck's 2025 annual report gives the dates plainly. Keytruda's key patents expire in 2028 in the United States and China, 2031 in the main European markets and 2032–33 in Japan. Two further US patents run to 2029 because of extensions, and Merck expects to have to defend them in court. Biosimilar copies are already in development by at least a dozen companies, including Amgen, Samsung Bioepis, Celltrion, Sandoz and Formycon, and have launched in a few small markets such as Argentina; Merck expects their effect to be immaterial in 2026.

The second blow comes from Washington. Under the Inflation Reduction Act, Medicare negotiates prices for the most expensive medicines. Keytruda was not chosen for 2028, but Merck's latest 10-Q says it expects Keytruda to be selected in 2027, with the government price taking effect on 1 January 2029, and a pending rule may sweep the new under-the-skin version in with it. Merck lost its constitutional challenge to the programme on 24 August 2026. In Germany, a health-spending reform effective mostly from January 2027 will, in Merck's words, exert significant downward pressure on revenue there, "particularly with respect to KEYTRUDA".

The third blow is competition. A new class of drugs that combine a PD-1 blocker, Keytruda's mechanism, with a VEGF blocker in one molecule is being tested against Keytruda in lung cancer, Keytruda's largest use. Summit and Akeso's ivonescimab has an FDA decision date of 14 November 2026, the day of this report, for a narrower use; Pfizer, BioNTech and Bristol Myers have their own versions in late trials. If lung-cancer treatment moves to these drugs between 2028 and 2030, it would arrive at exactly the moment the copies and the government price do.

What management says. The chief financial officer, Caroline Litchfield, said on 29 October 2026 to expect "a shallow dip in revenues" after the IRA price and the loss of exclusivity, followed by a quick return to growth, with earnings "probably" mirroring that. Analysts' consensus, which is thin beyond 2027, shows no dip at all: earnings per share of about $9.9 in 2028, 2029 and 2030. That is the most optimistic reading of the cliff we have seen. Our valuation in Part XI does not assume it.

In May 2025 the market was pricing Merck as if the cliff would be a fall. It had reasons: the China Gardasil collapse, the US government's threats on prices and tariffs, and two years to go before Keytruda's patent ended. At $73.47 the shares traded at about 8 times earnings, the lowest in at least a decade.

Since then three things happened. The policy threats were settled by a deal that cost Merck less than feared and postponed tariffs. A string of trial results went right: the melanoma vaccine, Lipfendra, sac-TMT in endometrial cancer, Winrevair's growth. And investors moved back into pharmaceutical stocks as a group. Meanwhile, the earnings did not move: Merck earned $8.98 a share before one-off items in 2025 and will earn about $8.85 in 2026 by the same measure. The share price doubled because the market changed its mind about what happens after 2028, not because anything happened before it. That is a legitimate reason for a price to move. It is also the kind of move that reverses when a trial fails.

How profitable. In 2025 Merck's pharmaceutical segment earned $45.8bn of segment profit on $58.1bn of sales, a 79% margin, before most of the research budget, which Merck keeps at the centre. Animal Health earned $2.1bn on $6.4bn. If Keytruda earns only the average margin of the pharmaceutical segment, it accounts for about 52% of total segment profit; a mature, high-priced biologic probably earns more than the average. Either way, the 2028–29 cliff is about more than half of the company's earning power.

The vaccine lesson. Gardasil shows how fast a franchise can change. It sold $8.6bn in 2024 and $5.2bn in 2025, because sales to China, through the distributor Zhifei, stopped: demand fell, a Chinese-made nine-valent vaccine was approved in May 2025, and Merck shipped nothing to China in 2025. A revised contract was signed in April 2026; revenue from it will be immaterial this year. In the US, a federal court in March 2026 paused a government recommendation for a single dose of HPV vaccine instead of two; Merck says any effect is not material. Vaccines depend on public-health policy, and public-health policy in the US has become less predictable.

A warning in the third quarter. About $550m of the quarter's $18.1bn of sales came from the timing of US purchases, which Merck expects to unwind in the fourth quarter. The third-quarter beat was real but smaller than it looked.

A drug company's moat is unusual: the strongest part of it, the patent, is a legal monopoly with a known expiry date. What lasts is the machinery that produces the next patent. We test both.

Our verdict is a moat of 7, of the intangible-asset kind, and its trend over the next five years is narrowing by construction: half the company's profit sits behind patents that end in 2028–2033. What would widen it again is the replacement portfolio becoming as broad as Keytruda's indications. Merck's scientists have done this before; Keytruda itself came to Merck with Schering-Plough in 2009, from research begun at Organon, and was nearly licensed away. The record of big pharmaceutical companies in replacing their largest product without a fall in earnings is mixed: AbbVie, after Humira, is the best recent example, and even AbbVie's earnings dipped for two years.

Robert Davis, Merck's chief executive, set out the case on 3 February 2026: "line of sight to over $70 billion of potential commercial opportunity by the mid-2030s", $20bn more than a year earlier and more than double the $35bn of Keytruda sales that analysts expect at the 2028 peak. Davis added that the figure was not risk-adjusted: it is what the pipeline could sell if everything works. On 29 October Davis said Merck expects to raise it. An owner should treat a non-risk-adjusted figure as an upper bound. In pharmaceutical research, a drug entering phase 3 trials has historically had something like a one-in-two chance of approval, and an approved drug rarely reaches the peak sales forecast for it.

Our reading. Of the eight largest purchases since 2021, about $51bn in total, only Acceleron has clearly earned its price so far; three others have products on the market that must still grow into their cost, and the rest are trials. That is not a criticism peculiar to Merck: buying late-stage science is how large drug companies replace their patents, and the prices are high because everyone is doing it at once. It does mean that a large part of today's share price rests on outcomes that have not yet happened. A Buffett-style test is whether the $60bn will produce something like $5–6bn a year of after-tax profit in the 2030s. Today, one deal clearly passes.

The pattern. Keytruda's competitors used to be other PD-1 drugs that it outsold by a wide margin. The threat now is different in kind: a drug that does what Keytruda does and something more, tested head-to-head against Keytruda in lung cancer. Merck's response is to make its own bispecific and to move Keytruda's users to combinations built on Qlex before the copies arrive. Whether that works will be known only between 2027 and 2030.

Across our board. We said AbbVie was "fairly priced — start small, add near $235" in September: AbbVie had already come through the Humira cliff and showed its replacements selling. Pfizer, in August, was "not a melting ice cube, not a compounder either". Merck sits between them: further from its cliff than Pfizer, but with the cliff still ahead rather than behind, as AbbVie's now is. At 16.5 times earnings, it is priced closer to AbbVie than to Pfizer.

Integrity. Merck's disclosure is thorough: its 10-Q gives the exact patent dates, the IRA expectations, every lawsuit and every injunction, and its management names the dip ahead rather than hiding it. Two things we would change. Merck's "non-GAAP" earnings, unlike most companies', include the cost of research bought in acquisitions, which made 2026 guidance of $2.77–2.82 a share look like a collapse when the business was flat; Merck is right to include it, because buying research is a real cost, but owners need both figures. And the "$70 billion" pipeline number, not risk-adjusted, is a sales pitch, not a forecast.

Capital allocation. Merck spends its cash in four directions: the dividend (about $8.4bn a year), buybacks ($3bn planned in 2026, $5.1bn left under the current authorisation), research, and deals. From 2016 to 2025 it bought back $31.2bn of stock and reduced its diluted share count by about 10%. Its debt rose from $24.8bn in 2016 to $53.8bn in September 2026, mostly to pay for the deals. The question for an owner is whether paying 10 to 15 times sales for companies with one promising medicine is a good use of Keytruda's cash. Acceleron says yes; the rest have yet to answer.

Governance. The roles of chairman and chief executive are combined, as at many large US companies. The board added a director on 1 November 2026. Pay is moderate for the sector; performance shares are linked half to earnings and half to relative shareholder return. No insider bought shares in the open market in the last twelve months.

At 30 September 2026 Merck had $53.8bn of debt and $11.3bn of cash and investments: net debt of about $42.5bn. On reported earnings, which include $14.7bn of research bought with Cidara and Terns and written off at once, that is about 3.0 times EBITDA; adding those one-off charges back, about 1.5 times. Merck's new bonds in May 2026 were rated Aa3 by Moody's and A+ by S&P, both stable. Goodwill and other intangible assets ($46.4bn) exceed shareholders' equity ($44.9bn), so tangible equity is slightly negative, which is normal for a company that has bought its pipeline. Interest cost rose to $1.55bn in the first nine months of 2026, from $0.95bn a year earlier.

Two kinds of earnings. Merck reports earnings under accounting rules (GAAP) and its own "non-GAAP" measure, which removes amortisation and restructuring but, unusually, keeps the cost of research bought in acquisitions. In 2026 that cost is $6.05 a share. So non-GAAP EPS for 2026 is guided at $2.77–2.82, GAAP lower still, and the underlying business earns about $8.85. All three numbers are true. Only the last tells you what the business earns in a year; the first two tell you what management spent on the future.

Merck pays $0.85 a quarter, $3.40 a year, a yield of 2.3%, the lowest at any year-end in the last decade (the range was 2.5–3.5%). Dividends declared per share have risen every year since 2011, fifteen years in a row, after four flat years from 2007 to 2010. The raises have been modest and slowing: +5.8% in November 2022, then +5.5%, +5.2% and +4.9%. Last year's was announced on 18 November; this year's had not been announced when we wrote. Over ten years the dividend has grown about 6% a year.

Our reading. This is the safest dividend of the three companies we analysed this week, with the lowest yield. Merck pays out less than 40% of what its business earns, so even a deep Keytruda cliff would not threaten it. The reason the yield is only 2.3% is the price, not the dividend: at the May 2025 low the same $3.24 a year then paid yielded 4.4%. An owner who bought then has done very well. A buyer today is paying for the pipeline, and the dividend is a small part of the return.

On today's earnings, Merck looks reasonably priced: 16–17 times, a little above the large-pharma average. The trouble is that today's earnings are the last of their kind. Valuing Merck means valuing 2030, after the cliff, and asking what the market will pay then.

Our method. We estimate earnings per share in 2030 (Merck's non-GAAP measure, before one-off acquired research), apply an exit multiple, discount 4.1 years at 10%, and add the dividends received on the way (about $12 a share in present value).

The consensus case moves to about $105 at 14 times and $132 at 18 times; to $104 at $8.50 of 2030 EPS and $131 at $11.00. What is priced in? At $147.76, Merck must earn about $12.50 a share in 2030 if the market then pays 16 times, or about $11.15 at 18 times: 26% to 42% more than today's underlying earnings, on the far side of losing the patent on half the company. The price sits at our best case. It assumes the pipeline wins and the market rewards it.

Margin of safety. We cannot estimate Merck's 2030 earnings with the confidence our method asks for; this is the "uncertainty present" tier at best, which asks for a 40–50% discount to a central value. We will not ask for that, because it would mean a price Merck reached only in the panic of May 2025. But we will ask to pay no more than our central case for a business whose next five years depend on trials, courts and Washington. We would add near $115: a 3.0% yield, about 13 times 2027 consensus, and roughly where the shares were in February 2026, nine months ago.